{
  "schemaVersion": "attractivenesstest-calibration-benchmark/v1",
  "engine": {
    "engineVersion": "0.2.0",
    "referenceVersion": "0.2.0-frll",
    "weightsVersion": "fixed-0.1.0"
  },
  "dataset": {
    "name": "Face Research Lab London Set",
    "doi": "10.6084/m9.figshare.5047666.v5",
    "license": "CC BY 4.0",
    "ratersInFile": 2513,
    "ratersPerFaceMedian": 2513,
    "neutralImages": 102,
    "smilingImages": 102
  },
  "sample": {
    "scoredNeutralFaces": 101,
    "male": 53,
    "female": 48,
    "neutralRejectedByIssue": {
      "YAW_TOO_LARGE": 1
    }
  },
  "agreement": {
    "note": "Geometry = 0.35 Harmony + 0.25 Dimorphism + 0.2 Angularity (Misc is constant in the free score and cancels in ranks). Human score = mean rating z within sex.",
    "inSampleSpearman": 0.446,
    "inSampleBootstrap95": {
      "p2_5": 0.268,
      "median": 0.447,
      "p97_5": 0.597
    },
    "bootstrapResamples": 2000,
    "heldOutSplitHalfSpearman": {
      "p2_5": 0.153,
      "median": 0.356,
      "p97_5": 0.522
    },
    "heldOutSplits": 500,
    "heldOutParitySplits": [
      0.27,
      0.35
    ],
    "bySexInSample": {
      "male": {
        "n": 53,
        "spearmanInSample": 0.45
      },
      "female": {
        "n": 48,
        "spearmanInSample": 0.45
      }
    },
    "humanSplitHalfSpearman": {
      "p2_5": 0.997,
      "median": 0.998,
      "p97_5": 0.999
    },
    "humanSplits": 500,
    "singleRaterVsRestSpearman": {
      "raters": 2501,
      "excludedConstantRaters": 12,
      "p25": 0.45,
      "median": 0.555,
      "p75": 0.638,
      "shareOfRatersBelowGeometryHeldOutMedian": 0.133,
      "shareOfRatersBelowGeometryInSample": 0.244
    }
  },
  "level": {
    "pslMidpoint": {
      "mean": 4.71,
      "p10": 3.75,
      "median": 4.75,
      "p90": 5.45,
      "min": 3.25,
      "max": 5.95
    },
    "pslRangeWidth": 0.5,
    "humanRating1to7": {
      "mean": 3.01,
      "p10": 2.13,
      "median": 2.95,
      "p90": 3.95,
      "min": 1.56,
      "max": 5.68
    },
    "shareOfScaleMean": {
      "psl": 0.529,
      "human": 0.335
    },
    "shareAboveScaleMidpoint": {
      "psl": 0.644,
      "human": 0.089
    }
  },
  "quintiles": [
    {
      "quintile": 1,
      "n": 20,
      "meanPslMidpoint": 3.78,
      "meanHumanRating1to7": 2.56,
      "meanHumanZ": -0.54
    },
    {
      "quintile": 2,
      "n": 20,
      "meanPslMidpoint": 4.41,
      "meanHumanRating1to7": 2.73,
      "meanHumanZ": -0.48
    },
    {
      "quintile": 3,
      "n": 21,
      "meanPslMidpoint": 4.78,
      "meanHumanRating1to7": 3.18,
      "meanHumanZ": 0.08
    },
    {
      "quintile": 4,
      "n": 20,
      "meanPslMidpoint": 5.04,
      "meanHumanRating1to7": 3.07,
      "meanHumanZ": 0.33
    },
    {
      "quintile": 5,
      "n": 20,
      "meanPslMidpoint": 5.53,
      "meanHumanRating1to7": 3.51,
      "meanHumanZ": 0.61
    }
  ],
  "expression": {
    "pairs": 101,
    "qualityGate": {
      "passed": 68,
      "rejected": 33,
      "rejectedByIssue": {
        "YAW_TOO_LARGE": 1,
        "MOUTH_OPEN": 32,
        "PITCH_TOO_LARGE": 2
      }
    },
    "gatedPassedOnly": {
      "n": 68,
      "meanDiff": -0.23,
      "meanAbsDiff": 0.38,
      "spearmanNeutralVsSmiling": 0.74
    },
    "ungated": {
      "n": 101,
      "meanDiff": -0.31,
      "meanAbsDiff": 0.45,
      "medianAbsDiff": 0.4,
      "p90AbsDiff": 0.9,
      "shareAbsDiffAtLeast0_5": 0.426,
      "spearmanNeutralVsSmilingGeometry": 0.66
    },
    "note": "Ungated = the same landmark normalisation and measurements applied even when the live quality gate would refuse the photo (the site would not score these)."
  },
  "noise": [
    {
      "sigmaShareOfFaceWidth": 0.0025,
      "repsPerFace": 100,
      "medianPerFaceSdPsl": 0.187,
      "p90PerFaceSdPsl": 0.238,
      "meanAbsDeltaPsl": 0.181,
      "meanSignedDeltaPsl": -0.089,
      "shareAbsDeltaAtLeast0_5": 0.0559,
      "medianRankCorrWithUnperturbed": 0.926,
      "qualityGatePassRate": 1
    },
    {
      "sigmaShareOfFaceWidth": 0.005,
      "repsPerFace": 100,
      "medianPerFaceSdPsl": 0.286,
      "p90PerFaceSdPsl": 0.331,
      "meanAbsDeltaPsl": 0.374,
      "meanSignedDeltaPsl": -0.31,
      "shareAbsDeltaAtLeast0_5": 0.3428,
      "medianRankCorrWithUnperturbed": 0.806,
      "qualityGatePassRate": 1
    },
    {
      "sigmaShareOfFaceWidth": 0.01,
      "repsPerFace": 100,
      "medianPerFaceSdPsl": 0.345,
      "p90PerFaceSdPsl": 0.404,
      "meanAbsDeltaPsl": 0.84,
      "meanSignedDeltaPsl": -0.822,
      "shareAbsDeltaAtLeast0_5": 0.7723,
      "medianRankCorrWithUnperturbed": 0.58,
      "qualityGatePassRate": 0.998
    },
    {
      "sigmaShareOfFaceWidth": 0.02,
      "repsPerFace": 100,
      "medianPerFaceSdPsl": 0.336,
      "p90PerFaceSdPsl": 0.373,
      "meanAbsDeltaPsl": 1.475,
      "meanSignedDeltaPsl": -1.472,
      "shareAbsDeltaAtLeast0_5": 0.9479,
      "medianRankCorrWithUnperturbed": 0.287,
      "qualityGatePassRate": 0.899
    }
  ],
  "seeds": {
    "bootstrap": 20261021,
    "heldOut": 7,
    "humanSplit": 99,
    "noise": 314159
  },
  "generatedAt": "2026-10-21T02:56:51.040Z",
  "benchmarkVersion": "2026-10-21-v1",
  "dataBoundary": "Aggregate numbers only. No image, landmark or per-face score from the Face Research Lab London Set is published."
}
